64

However, a so-called semiquantitative model reproduces the processes somewhat more

accurately. Starting from the Boolean network, differential equations, e.g. exponential

functions, are linked in such a way that they reproduce this logic, but by means of the

mathematical transformation between the completely switched-on or switched-off state,

they lay a compensation curve (“interpolate”). In order to reproduce the logic in the net­

work correctly, the software SQUAD, for example, creates chained exponential terms

(uses exponential function), which also take into account the “and”, “or” and “not”. It

reads networks written with CellDesigner e.g. as SBML format and requires a Windows

XP or Linux operating system. These limitations no longer apply to the Jimena software

(Karl and Dandekar 2013a). It runs platform-independently using Java and can read YeD

files, among others, but also various versions of CellDesigner. Surprisingly, this way I also

get all order relations in the model correctly, i.e. which receptor is excited before which

one and which link in a signal chain is activated earlier or later. In most cases, the mole­

cules close to the receptor are excited first, followed by the later, mediating proteins. If the

topology (structure) of the model provides for a feedback loop, this can then return the

signal to the beginning of the signal chain, either inhibiting (negative feedback) or activat­

ing (positive feedback, sometimes also called “feedforward loop”).

This brings us to another important point. The software can only simulate correctly

what is also reproduced correctly in the network. This means that a period of constant test­

ing and trial and error begins until the simulation reproduces the correct sequence of

events in this signal network as faithfully as possible.

Since this is a semi-quantitative model, the next step is to normalize the different units

of the model according to the experimental data. This means that the typical times of the

signal cascade, receptor excitation, phosphorylation of kinases, etc. are determined

(so-­called data-driven modeling). Hundreds of biological problems have already been

simulated in this way in recent years. The Boolean semiquantitative model is therefore

quite popular in biology, because one can begin to describe the biological system with

relatively little information, and then step by step learn more and more about the model

through simulations and experiments.

If so much data is put into the model, one can of course wonder what new insights the

model can bring out. But it is the case that a few experiments are sufficient to normalize

the model and to qualitatively confirm the correctness of all links (correct stimulus

response and sequence). With the model, I can now predict the outcome for all times and

all signal and switching sequences that are possible in the network.

For example, we used this approach to simulate the behavior of lung carcinoma

(Stratmann et al. 2014; Göttlich et al. 2016a) and colon carcinoma cells (Baur et al. 2019)

and then tested through new combinations and options for therapies in addition to standard

therapies.

With regard to the Erk signaling network, the interesting thing was that through the

bioinformatic model we can mimic new approaches to treating heart failure (Brietz et al.

2016a), such as the negative feedback loop through Rkip or the approach of using

5  Systems Biology Helps to Discover Causes of Disease